We Media" have enabled netizens to express their opinions in a freer and more convenient way. Netizens' speech tends to be deviated after group discussion, resulting in online public opinion group polarization and irrational behaviours. Here, a quantitative index of Microblog public opinion group polarization has been established based on the sentiment analysis of the comment text, the popularity of comments and the interaction of users' @. Besides, the effectiveness of such index has been verified by simulation experiment, aiming to provide decision-making thinking and method guidance for Microblog public opinion group polarization determination.On the Internet, some marketing accounts of "We Media" are quite easy to push with fragmental verbal information of a certain topic to the extreme within a short time. Under the environment of high public sentiment, netizens are likely to be coerced by simple arbitrary sentimental languages. In China, the public's urgent needs for high quality higher education resources have made news events in colleges and universities to win the public's close attention. Most colleges and universities attach great importance to their image-building. In recent years, however, occasional negative news in colleges and universities has developed into online public opinion events with the boost of the Internet. It has had a huge impact on the reputation of the school and brought a great challenge to the image crisis management in colleges and universities. If Internet rumours can be identified in time, we can communicate with netizens in time to eliminate misunderstandings,more time will be earned for crisis management. Thus, the school's reputation can be maintained. In this paper, related causes for public opinion group polarization will be analysed. On this basis, we will also explore a group polarization behaviour pattern and establish a quantitative index for group polarization. With public opinion events in a Chinese university as an example, the effectiveness of the method in this paper will be verified as well.